collectd (1) - Linux Man Pages
collectd: System statistics collection daemon
collectd - System statistics collection daemon
DESCRIPTIONcollectd is a daemon that receives system statistics and makes them available in a number of ways. The main daemon itself doesn't have any real functionality apart from loading, querying and submitting to plugins. For a description of available plugins please see ``PLUGINS'' below.
OPTIONSMost of collectd's configuration is done using using a configfile. See collectd.conf(5) for an in-depth description of all options.
- -C <config-file>
- Specify an alternative config file. This is the place to go when you wish to change collectd's behavior. The path may be relative to the current working directory.
- Test the configuration only. The program immediately exits after parsing the config file. A return code not equal to zero indicates an error.
- Test the plugin read callbacks only. The program immediately exits after invoking the read callbacks once. A return code not equal to zero indicates an error.
- -P <pid-file>
- Specify an alternative pid file. This overwrites any settings in the config file. This is thought for init-scripts that require the PID-file in a certain directory to work correctly. For everyday-usage use the PIDFile config-option.
- Don't fork to the background. collectd will also not close standard file descriptors, detach from the session nor write a pid file. This is mainly thought for 'supervising' init replacements such as runit. If using upstart or systemd though, starting with version 5.5.0 collectd is able to notify these two init replacements, and does require forking to the background for process supervision. The contrib/ directory has sample upstart and systemd configuration files.
- Output usage information and exit.
PLUGINSAs noted above, the real power of collectd lies within it's plugins. A (hopefully complete) list of plugins and short descriptions can be found in the README file that is distributed with the sourcecode. If you're using a package it's a good bet to search somewhere near /usr/share/doc/collectd.
There are two big groups of plugins, input and output plugins:
- Input plugins are queried periodically. They somehow acquire the current value of whatever they where designed to work with and submit these values back to the daemon, i. e. they ``dispatch'' the values. As an example, the "cpu plugin" reads the current cpu-counters of time spent in the various modes (user, system, nice, ...) and dispatches these counters to the daemon.
- Output plugins get the dispatched values from the daemon and does something with them. Common applications are writing to RRD-files, CSV-files or sending the data over a network link to a remote box.
Of course not all plugins fit neatly into one of the two above categories. The "network plugin", for example, is able to send (i. e. ``write'') and receive (i. e. ``dispatch'') values. Also, it opens a socket upon initialization and dispatches the values when it receives them and isn't triggered at the same time the input plugins are being read. You can think of the network receive part as working asynchronous if it helps.
In addition to the above, there are ``logging plugins''. Right now those are the "logfile plugin" and the "syslog plugin". With these plugins collectd can provide information about issues and significant situations to the user. Several loglevels let you suppress uninteresting messages.
Starting with version 4.3.0 collectd has support for monitoring. This is done by checking thresholds defined by the user. If a value is out of range, a notification will be dispatched to ``notification plugins''. See collectd.conf(5) for more detailed information about threshold checking.
Please note that some plugins, that provide other means of communicating with the daemon, have manpages of their own to describe their functionality in more detail. In particular those are collectd-email(5), collectd-exec(5), collectd-perl(5), collectd-snmp(5), and collectd-unixsock(5)
SIGNALScollectd accepts the following signals:
- SIGINT, SIGTERM
- These signals cause collectd to shut down all plugins and terminate.
- This signal causes collectd to signal all plugins to flush data from internal caches. E. g. the "rrdtool plugin" will write all pending data to the RRD files. This is the same as using the "FLUSH -1" command of the "unixsock plugin".
AUTHORFlorian Forster <octo [at] collectd.org>